A Voting strategy for N-Version Program based on Fuzzy Clustering

The thesis study the consensus question about float data in N-Version Programming, and put forward a voting strategy based on dynamic fuzzy cluster. At first, the strategy process the output from multi-time operations with standardization, and then analyze the degree of membership between each data, so produce the λ-cut set by dynamic fuzzy cluster method according to the degree of membership, the ultimate output is decided by the consensus of majority number of members. The strategy can eliminate the incorrect or error data, judge the consistency between data correctly, and then acquire the correct output.